A regional early warning model of geological hazards based on big data of real-time rainfall

• Published in: Natural hazards (Dordrecht) Vol. 116; no. 3; pp. 3465 - 3480
• Main Authors: Zhao, Weidong, Cheng, Yunyun, Hou, Jie, Chen, Yihua, Ji, Bin, Ma, Lei
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01-04-2023; Springer Nature B.V
• Subjects: Accuracy; Algorithms; Big Data; Civil Engineering; Earth and Environmental Science; Earth Sciences; Environmental Management; False alarms; Forecast accuracy; Geological hazards; Geology; Geophysics/Geodesy; Geotechnical Engineering & Applied Earth Sciences; Hazard mitigation; Hourly rainfall; Hydrogeology; Hydrologic data; Impact damage; Mathematical models; Mitigation; Modelling; Natural Hazards; Original Paper; Precipitation; Probability theory; Rain; Rainfall; Rainfall data; Rainfall forecasting; Real time; Similarity; Early warning model; Rainfall big data; Hourly rainfall series; Geological hazards; Dynamic time warping
• ISSN: 0921-030X; 1573-0840
• DOI: 10.1007/s11069-023-05819-z

The warning accuracy, false alarm rate and timeliness of regional geological hazard early warning models (GHEWMs) have an important impact on significantly reducing the damage caused by geological hazards. Most of the existing regional GHEWMs are based on forecast rainfall. Due to the influence of rainfall forecast accuracy and other factors, its early warning accuracy, false alarm rate and timeliness are still difficult to meet the needs of engineering applications such as disaster avoidance, mitigation and prevention of geological hazards. Therefore, this paper proposes a regional GHEWM based on the hourly rainfall series (HRS) of real-time automatic rainfall stations. Based on the data of 689 geological hazards that have occurred in Huangshan City from 2018 to 2021 and the corresponding rainfall data of automatic rainfall stations, the model uses the dynamic time warping (DTW) algorithm on the Spark big data platform to extract the historical HRS of each geological hazard and calculates the highest similarity between it and the current HRS in parallel. By coupling the probability of occurrence of geological hazards and the highest similarity of the above-mentioned HRS, a regional GHEWM based on real-time rainfall big data is finally constructed. The research results show that the model's early warning accuracy reaches 85%, and the false alarm rate is only 15%, which can predict the possibility of geological hazards after the next 3 h.